Merge pull request #713 from htm-community/examples_fix

Python examples fixes

VERSION

@@ -1 +1 @@
-v2.0.16
+v2.0.22

bindings/py/tests/algorithms/sdr_classifier_test.py

@@ -18,6 +18,7 @@
 
 """ Unit tests for Classifier & Predictor classes. """
 
+import math
 import numpy
 import pickle
 import random
@@ -131,14 +132,13 @@ def testComputeInferOrLearnOnly(self):
     inp.randomize( .3 )
 
     # learn only
-    with self.assertRaises(RuntimeError):
-      c.infer(pattern=inp) # crash with not enough training data.
+    prediction = c.infer(pattern=inp)[1]
+    self.assertTrue(prediction == []) # not enough training data -> []
     c.learn(recordNum=0, pattern=inp, classification=4)
-    with self.assertRaises(RuntimeError):
-      c.infer(pattern=inp) # crash with not enough training data.
+    self.assertTrue(c.infer(pattern=inp)[1] == []) # not enough training data.
     c.learn(recordNum=2, pattern=inp, classification=4)
     c.learn(recordNum=3, pattern=inp, classification=4)
-    c.infer(pattern=inp) # Don't crash with not enough training data.
+    self.assertTrue(c.infer(pattern=inp)[1] != []) # Don't crash with enough training data.
 
     # infer only
     retval1 = c.infer(pattern=inp)

py/htm/examples/hotgym.py

@@ -10,7 +10,7 @@
 from htm.encoders.date import DateEncoder
 from htm.bindings.algorithms import SpatialPooler
 from htm.bindings.algorithms import TemporalMemory
-from htm.algorithms.anomaly_likelihood import AnomalyLikelihood
+from htm.algorithms.anomaly_likelihood import AnomalyLikelihood #FIXME use TM.anomaly instead, but it gives worse results than the py.AnomalyLikelihood now
 from htm.bindings.algorithms import Predictor
 
 _EXAMPLE_DIR = os.path.dirname(os.path.abspath(__file__))
@@ -158,18 +158,20 @@ def main(parameters=default_parameters, argv=None, verbose=True):
     tm_info.addData( tm.getActiveCells().flatten() )
 
     # Predict what will happen, and then train the predictor based on what just happened.
-    pdf = predictor.infer( count, tm.getActiveCells() )
+    pdf = predictor.infer( tm.getActiveCells() )
     for n in (1, 5):
       if pdf[n]:
         predictions[n].append( np.argmax( pdf[n] ) * predictor_resolution )
       else:
         predictions[n].append(float('nan'))
-    predictor.learn( count, tm.getActiveCells(), int(consumption / predictor_resolution))
 
     anomalyLikelihood = anomaly_history.anomalyProbability( consumption, tm.anomaly )
     anomaly.append( tm.anomaly )
     anomalyProb.append( anomalyLikelihood )
 
+    predictor.learn(count, tm.getActiveCells(), int(consumption / predictor_resolution))
+
+
   # Print information & statistics about the state of the HTM.
   print("Encoded Input", enc_info)
   print("")
@@ -189,6 +191,7 @@ def main(parameters=default_parameters, argv=None, verbose=True):
   # Calculate the predictive accuracy, Root-Mean-Squared
   accuracy         = {1: 0, 5: 0}
   accuracy_samples = {1: 0, 5: 0}
+
   for idx, inp in enumerate(inputs):
     for n in predictions: # For each [N]umber of time steps ahead which was predicted.
       val = predictions[n][ idx ]

py/htm/examples/sp/sp_tutorial.py

@@ -103,8 +103,7 @@ def corruptSDR(sdr, noiseLevel):
 # input. As well, the histogram will show the scores of those columns
 # that are chosen to build the sparse representation of the input.
 
-sp.compute(inputSDR, False, outputSDR)
-overlaps = sp.getOverlaps()
+overlaps = sp.compute(inputSDR, False, outputSDR)
 activeColsScores = []
 for i in outputSDR.sparse:
   activeColsScores.append(overlaps[i])
@@ -240,10 +239,12 @@ def corruptSDR(sdr, noiseLevel):
 # This is the number of times that we will present the input vectors to the SP
 epochs = 30
 
+overlapsUntrained = overlaps
+
 for _ in range(epochs):
   for i in range(numExamples):
     # Feed the examples to the SP
-    sp.compute(inputVectors[i], True, outputColumns[i])
+    overlaps = sp.compute(inputVectors[i], True, outputColumns[i])
 
 print("")
 print("---------------------------------")
@@ -254,8 +255,7 @@ def corruptSDR(sdr, noiseLevel):
 print("---------------------------------")
 print("")
 
-plt.plot(sorted(overlaps)[::-1], label="Before learning")
-overlaps = sp.getOverlaps()
+plt.plot(sorted(overlapsUntrained)[::-1], label="Before learning")
 plt.plot(sorted(overlaps)[::-1], label="After learning")
 plt.axvspan(0, len(activeColsScores), facecolor="g", alpha=0.3, label="Active columns")
 plt.legend(loc="upper right")

src/htm/algorithms/SDRClassifier.cpp

@@ -47,8 +47,10 @@ void Classifier::initialize(const Real alpha)
 PDF Classifier::infer(const SDR & pattern) const {
   // Check input dimensions, or if this is the first time the Classifier is used and dimensions
   // are unset, return zeroes.
-  NTA_CHECK( dimensions_ != 0 )
-    << "Classifier: must call `learn` before `infer`.";
+  if( dimensions_ == 0 ) {
+    NTA_WARN << "Classifier: must call `learn` before `infer`.";
+    return PDF(numCategories_, std::nan("")); //empty array []
+  }
   NTA_ASSERT(pattern.size == dimensions_) << "Input SDR does not match previously seen size!";
 
   // Accumulate feed forward input.

src/htm/algorithms/SDRClassifier.hpp

@@ -128,6 +128,7 @@ class Classifier : public Serializable
    * @param pattern: The SDR containing the active input bits.
    * @returns: The Probablility Distribution Function (PDF) of the categories.
    *           This is indexed by the category label.
+   *           Or empty array ([]) if Classifier hasn't called learn() before. 
    */
   PDF infer(const SDR & pattern) const;